RCA
Elite Member
- Joined
- Jun 27, 2017
- Professional Status
- Certified General Appraiser
- State
- California
There are several people continually downloading some of my apps, in particular earthUI. Neither GitHub nor CRAN tells me who the downloaders are. So, if any are on this forum, the news is that, for the time being, I am stopping further development on the current version earthUI. In fact, I removed some of the latest changes and rolled back to version 0.11. The latest official release is 0.9.0.
Development is progressing on the earthUI clone, "earthUI2," which is currently a private repository. The changes are somewhat radical and geared towards supporting analysts who wind up creating complex histories in their workflows, rerunning with different parameter sets, or updating their data (adding more information or collecting more data). As you may know, I have created my own version of "vProlog," which is based on the embeddable "Trealla Prolog." Anyway, short story, for the time being, things are getting complicated. I want to keep everything R/Shiny/vProlog, but I may switch from Shiny to Angular. Claude argues against moving away from the pure R UI. We will see.
I am extracting all available information in the MLS data for valuation. That includes public and agent remarks, fields with feature lists, and so on. This results in long lists of variables to manage. The simplest possible way to do this has to be designed. So, it is more work. It is a question of managing complexity as simply as possible.
But my system has already processed the remarks and string feature fields for more than 2M sales transactions across the SF Bay Area. It can break down about 50% of the residual based on textual information alone. The other 10% has been taken care of by the appraiser at present. But once photos are pulled into analysis, it will start getting close to a 100% explanation of comparable differences to the subject. Right now, I am working on Woodside, CA, which is certainly an interesting and varied area.
Development is progressing on the earthUI clone, "earthUI2," which is currently a private repository. The changes are somewhat radical and geared towards supporting analysts who wind up creating complex histories in their workflows, rerunning with different parameter sets, or updating their data (adding more information or collecting more data). As you may know, I have created my own version of "vProlog," which is based on the embeddable "Trealla Prolog." Anyway, short story, for the time being, things are getting complicated. I want to keep everything R/Shiny/vProlog, but I may switch from Shiny to Angular. Claude argues against moving away from the pure R UI. We will see.
I am extracting all available information in the MLS data for valuation. That includes public and agent remarks, fields with feature lists, and so on. This results in long lists of variables to manage. The simplest possible way to do this has to be designed. So, it is more work. It is a question of managing complexity as simply as possible.
But my system has already processed the remarks and string feature fields for more than 2M sales transactions across the SF Bay Area. It can break down about 50% of the residual based on textual information alone. The other 10% has been taken care of by the appraiser at present. But once photos are pulled into analysis, it will start getting close to a 100% explanation of comparable differences to the subject. Right now, I am working on Woodside, CA, which is certainly an interesting and varied area.
